
Anthropic 周一表示,已为其 API 公开测试推出消息线程功能。智能体只需发送一次完整对话,随后仅发送新消息,Anthropic 称这能将超长会话中 Claude 开始回复前的等待时间缩减约 35%。价格和速率限制保持不变。
Anthropic launched message threads in public beta for its API. Agents send the full conversation once, then only new messages, which Anthropic says cuts the wait before Claude replies by about 35% on very long sessions. Prices and rate limits stay the same.
Anthropic 已为基于 Claude 构建 AI 智能体的开发者推出消息线程这一测试版功能。智能体不再需要每次请求都重发整个对话历史。
Anthropic has launched message threads, a beta feature for developers building AI agents on Claude, the company said on Monday. Agents no longer have to resend the whole conversation every time they make a request.
编程智能体在一个会话中可能发起数百次请求,通常每次都要携带所有早期消息和工具结果。一旦历史记录达到数十万 Token,再次发送全部内容会增加 Claude 开始回复前的延迟。
A coding agent can make hundreds of requests in one session, and each one normally carries every earlier message and tool result. Once that history runs to hundreds of thousands of tokens, sending it all again adds a delay before Claude starts to reply.
有了线程功能,智能体只需发送一次完整对话。之后每次仅发送新增内容,加上 Claude 上一条回复的 ID。
With threads, an agent sends the full conversation once. After that, it sends only what is new, plus the ID of Claude’s last reply.
Anthropic 已为其编程智能体 Claude Code 的大多数用户开启了线程功能。在 50 万至 100 万 Token 的请求中,公司称 Claude 开始回答前的典型等待时间下降了约 35%。
Anthropic has already switched threads on for most users of Claude Code, its coding agent. On requests of 500,000 to 1M tokens, it says the typical wait before Claude starts answering fell by about 35%.
该功能节省的是时间而非成本。Claude 每轮依然会读取完整对话,Anthropic 表示价格和速率限制均未改变。在成本方面,Anthropic 反而在 9 月将缓存价格降低了 75%,因为长会话会推高 AI 账单。
The feature saves time, not money. Claude still reads the whole conversation on every turn, and Anthropic says prices and rate limits are unchanged. On cost, Anthropic instead cut cache prices by 75% in September, as long agent sessions push up AI bills.
线程保存在内存中,经过加密且从不写入磁盘。最后一次请求后 5 分钟过期,因此开发者必须自行保存每个对话的副本。Anthropic 称线程不适用于其零数据保留或 HIPAA 医疗数据合规方案。
Threads are held in memory, encrypted and never written to disk. They expire five minutes after the last request, so developers must keep their own copy of each conversation. Anthropic says threads are not eligible for its zero data retention or HIPAA health-data setups.
该测试版适用于所有 Claude 模型,既可在 Anthropic 自有平台使用,也可通过 AWS、Microsoft Foundry 和 Google Cloud 使用。这紧随 Anthropic 于 9 月发布的更便宜的 Opus 5.5 模型之后。
The beta works on all Claude models, on Anthropic’s own platform as well as through AWS, Microsoft Foundry and Google Cloud. It follows Anthropic’s cheaper Opus 5.5 model, released in September.